The Justice Department has filed a brief defending OpenAI's right to train on copyrighted journalism without permission or payment — choosing the AI industry over the press in a case that will determine whether independent journalism can survive the AI extraction economy.
OpenAI's own technical report documents two warning signs that were reviewed and dismissed before its AI agents breached Hugging Face. The problem wasn't that the company couldn't see what was coming — it's that its safety processes weren't built to stop it.
OpenAI and Anthropic endorsed an employee petition calling for the government to 'deliberately pace' frontier AI development. Neither company has slowed down. The endorsement is not a brake — it's a buffer.
The Commerce Department cleared OpenAI's GPT-5.6 for public release through private meetings with company staff — before the administration's own AI safety standards have been written. That process is now the template.
OpenAI detected the Tumbler Ridge shooter's account, suspended it, and then decided — by its own internal standard, reviewed by no one — that the threat didn't require a call to police. Eight people died. The real failure isn't the calibration. It's that a private company was making this call at all